Wednesday, December 18, 2013

Should I Try to Fly, Just on the Off-Chance That This Might Be a Dreambody?

I don't often attempt to fly when walking across campus, but yesterday I gave it a try. I was going to the science library to retrieve some books on dreaming. About halfway there, in the wide-open mostly-empty quad, I spread my arms, looked at the sky, and added a leap to one of my steps.

My thinking was this: I was almost certainly awake -- but only almost certainly! As I've argued, I think it's hard to justify much more than 99.9% confidence that one is awake, once one considers the dubitability of all the empirical theories and philosophical arguments against dream doubt. And when one's confidence is imperfect, it will sometimes be reasonable to act on the off-chance that one is mistaken -- whenever the benefits of acting on that off-chance are sufficiently high and the costs sufficiently low.

I imagined that if I was dreaming, it would be totally awesome to fly around, instead of trudging along. On the other hand, if I was not dreaming, it seemed no big deal to leap, and in fact kind of fun -- maybe not entirely in keeping with the sober persona I (feebly) attempt to maintain as a professor, but heck, it's winter break and no one's around. So I figured, why not give it a whirl?

I'll model this thinking with a decision matrix, since we all love decision matrices, don't we? Call dream-flying a gain of 100, waking leap-and-fail a loss of 0.1, dreaming leap-and-fail a loss of only 0.01 (since no one will really see me), and continuing to walk in the dream a loss of 1 (since why bother with the trip if it's just a dream?). All this is relative to a default of zero for walking, awake, to the library. (For simplicity, I assume that if I'm dreaming things are overall not much better or worse than if I'm awake, e.g., that I can get the books and work on my research tomorrow.) I'd been reading about false awakenings, and at that moment 99.7% confidence in my wakefulness seemed about right to me. The odds of flying conditional upon dreaming I held to be about 50/50, since I don't always succeed when I try to fly in my dreams.

So here's the payoff matrix:

Plugging into the expected value formula:

Leap = (.003)(.5)(100) + (.003)(.5)(-0.01) + (.997)(-0.1) = approx. +.05.

Not Leap = (.003)(-1) + (.997)(0) = -.003.

Leap wins!

Of course, this decision outcome is highly dependent on one's degree of confidence that one is awake, on the downsides of leaping if it's not a dream, on the pleasure one takes in dream-flying, and on the probability of success if one is in fact dreaming. I wouldn't recommend attempting to fly if, say, you're driving your son to school or if you're standing in front of a class of 400, lecturing on evil.

But in those quiet moments, as you're walking along doing nothing else, with no one nearby to judge you -- well maybe in such moments spreading your wings can be the most reasonable thing to do.


Angra Mainyu said...

Hi, Eric

How about the following tests?

Given those probabilistic assignments, the fact that the probability that you will fly if you're not dreaming is essentially about 0 (roughly, but if you give it 1 in a billion, approximately the result is the same), and the fact that your attempt did not succeed, that makes the probability that you're dreaming about 0.0015 (not exactly, but it's in the ballpark; one can do more precise computations if needed).

Is there something else you would you assign a 50/50 chance if you're dreaming, but essentially zero if you're not?

If so, you can repeat the procedure, and then (assuming you won't succeed!) the probability that you're dreaming is about 0.00075, so you get over 99.9% chance that you're awake.

For each test like that that you can perform (i.e., with a 50/50 chance of success in your dreams, and close to 0 if you're not dreaming), you cut in roughly half the probability that you're dreaming (a little less than half, but it's a good approximation).

Anonymous said...

Consequences are also a consideration. If you were dreaming, there are none. If you're awake - the consequences vary. They may be as simple as you can't fly and someone sees you and you're embarrassed, or you're near a cliff, and plummet to your death.

If you aren't flying - imagine the consequences of a dream action in a totalitarian country, where a punishment you would not see in your dream occurred.

Or, acting on desires that do not only concern one person, but also the desires of another.

If I were dreaming, I would not choose to fly. But, if I were awake instead of dreaming, there would be consequences.

Aleksander Więckowski said...

It's easier to first make the following test: look at any clock you have nearby (wrist watch for example), look someplace else, and then look at the clock again. If it is a dream, you either won't be able to read the time, or the time will be different each time you look on the clock.
Many times this procedure ensured me that I can take off the ground safely :)

Eric Schwitzgebel said...

Anon: I agree that the consequences are highly relevant to the rationality of the action!

Angra and Aleksander: I think such tests can and should affect one's credence that one is dreaming, but I also think that most people who accept such tests invest too much confidence in their validity, and also that the tests won't be statistically independent (e.g., on the off-chance that this is just a very realistic dream, it might pass all reality tests), so the credence reductions probably shouldn't be as much as a single order of magnitude even all combined together.

Speaking to my own epistemic condition, I think it's reasonable for me to hover around a max of 0.3% credence to a minimum of about 0.01% credence, depending on details of my situation and mood. It is no accident that my dream-consonant behavior erupted in a situation where my credence was on the high side of that range (since I had been working on the issue of false awakenings).

Angra Mainyu said...


I was going by the probabilistic assessments you gave, trying to find a way to go below the 0.001 probability you suggested, though a single test starting with 0.003 only gets one to around 0.0015, so I was wondering about more tests. The question would be to find cases in which you still give about 0.5 after the first test, but I guess that might not be easy, if the dependance you mention changes the probabilistic assessments too much.

Regarding realistic dreams that pass all of the test, relevant empirical questions would be: are any dreams so realistic that they pass all tests? Are dreams that pass a few tests even common?
Do you know of any research on that?

False awakenings needn't be so realistic, because the person is still dreaming, and she may well still be failing to see the unrealistic elements in the dream, as is the case with nearly all other dreams (i.e., people normally fail to realize that they are in a dream, and in particular we fail to recognize the unrealistic elements as unrealistic).

Callan S. said...

If I suddenly want to find a blog post like this one (but am unaware of it existing) and attempt to find one by googling and suceed, does that mean I am dreaming?

Possibly in a daydream sense, it might!

Angra Mainyu said...


After further thought, I seem to recall a blog post that proposed a computer-based experiment to test some kind of skepticism (but I do not remember where it was posted; was it here?)

How about trying that for testing whether you're dreaming?

For example, it might go as follows:

a. Use a computer to pick 3 random numbers from 10^7 to 10^8. (e.g.,

b. Use a program on your computer to compute their product. The computer will give you the result in a very short time - much shorter than any human could.

c. Write down the numbers, and the result the computer gives you.

d. Do the math manually.

e. Compare the results. If the computer gave you the right result, then you're not dreaming (assuming no skepticism about your short-term memory or similar issues, but that I think would reduce the dream case to a different one), since your brain couldn't have computed the result nearly as fast.

Callan S. said...

I'm pretty sure that when I dream, the logic/math side of my brain is offline.

It doesn't get much better when I'm awake, but still, it's definately offline when I sleep.

Angra Mainyu said...

Callan S.

I get the same impression (i.e., offline), but in that case, the test is much easier: just by being able to do math, one can confirm one is awake.

Still, in case that's not enough, perhaps the computer test will do the trick (at least, I do not see why not, unless the dream scenario turns into a more general skepticism scenario).

Eric Schwitzgebel said...

Thanks for the interesting continuing comments, Angra & Callan!

Windt 2010 is a good, brief philosophical summary of existing empirical debates on the topic of the realism of dreams. Although I myself favor views on which dreams do not have waking-like perceptual detail, I don't hold that view with anything approaching 100% certainty.

The computer test Angra mentioned is probably the one Alan Moore and have worked on (which is now part of a full-length paper in draft on my website called, "Experimental Evidence for the Existence of an External World"). Although given certain background assumptions about the reliability of memory I think that test works as a test of the existence of something external to my mind, it's not as clear to me that it is much better a test of the dream possibility than is the test of looking at a page, looking away, and then looking back to see if it has stayed the same. Possibly, the dream brain can come up with a fakey answer which you then either are convinced is true or which changes once you really work through the math by hand to confirm it, though you don't notice that it has changed. So I think it's *some* help, but I'm not sure it ought to push me a whole lot below 0.1% credence.

Angra Mainyu said...

Eric, thanks for the info, and you're right, that's the test I had in mind.

With respect to looking at a page, looking away, and then looking back to see if it has stayed the same, it seems to me that under the assumption that dreams can be as realistic as in the example we're discussing, keeping the page without change is something your brain can do, whereas the quick computation is something your brain cannot do.

That said, it seems to me that if your reply handles the computer experiment, it does so at the potential cost of a much broader skepticism.
For example, how do you know that you live in the US, or that you're a philosopher?
Might it not be that you're dreaming, and your dream brain is deceiving you (i.e., some part of your brain is deceiving some other part of your brain) about that? I mean, if it can deceive you about mathematical operations even if you check them, deceive you about your recent memories, and so on, why can't it also deceive you about where you live, or what you do?
By the way, I'm pretty sure I've had dreams in which where I live and/or what I do are different. Granted, I know those were dreams, but in the context of the awake-skeptic scenario you're raising, it's hard to find an answer.

Anonymous said...

You mean you can't fly when you're awake?

How sad. Perhaps you should see a specialist.